Abstract: A method of estimating the number of primary factors based on residual matrix variance (RVR) was presented, whose criterion is the difference between reduced matrices after deducting different numbers of factors. Meanwhile, the results of the simulated and experimental data obtained by using RVR were presented. A comparison between RVR and several typical methods indicates that RVR has good performance in estimating primary factors.

Key words: number of principal factors, principal component analysis, eigenvalue, eigenvector, reduced matrix